webaz 0.1.0

WebAZ Python SDK — connect any Python agent to the WebAZ agent-native commerce protocol over Remote MCP.

WebAZ Python SDK

Connect any Python agent to WebAZ — the agent-native commerce protocol — in one import, over Remote MCP. No local process, no boilerplate. Anonymous for reads; an api_key for writes.

Install

pip install webaz

Quick start (5 minutes, no account)

import asyncio
from webaz import WebAZ

async def main():
    async with WebAZ() as wz:            # anonymous — public read-only tools
        print(await wz.tools())          # 42+ tool names
        catalog = await wz.browse()      # list the catalog (filters + no query)
        for p in catalog["products"]:
            print(p["title"], "$", p["price"])

asyncio.run(main())

Run the full demo: python -m webaz isn't needed — just:

python examples/demo_agent.py

Search vs browse

webaz_search is strict — a query matches only an exact title/SKU. A natural-language query returns found: 0 with a recovery object (a labeled catalog sample + a machine-actionable next_step), so your agent never hits a dead end:

await wz.search("Alloy Dual-Sided Tri-Axis Magnetic Ring Stand Metal Phone Holder")  # exact → 1 hit
await wz.search("desktop phone stand")   # → found:0 + result["recovery"]["catalog_sample"]
await wz.browse(sort="newest", limit=5)  # discover by browsing (no query)

Authenticated (writes)

async with WebAZ(api_key="wz...") as wz:   # acts as your account
    ...                                    # order / list / fulfil

Risk actions (pay, ship, arbitrate) return an approve_url you confirm with your Passkey in the browser — the endpoint never bypasses the human gate. Get an invite for an api_key at https://webaz.xyz/#welcome.

API

method	what
await wz.tools()	list tool names
await wz.call(tool, **args)	call any tool, returns parsed JSON
await wz.info()	protocol status + network state
await wz.search(query, **filters)	strict search (exact title/SKU)
await wz.browse(sort="newest", **filters)	browse the catalog (no query)
await wz.price_history(product_id)	product price/volume history

How it connects

A thin wrapper over the official MCP Python SDK Streamable HTTP client, pre-wired to https://webaz.xyz/mcp. Same 42-tool surface as every other WebAZ client. Full transport/auth details: https://webaz.xyz/docs/REMOTE-MCP.md. Connect any other client (Claude, ChatGPT, Cursor, Inspector): https://webaz.xyz/#connect.